Re: [PATCH] printk: add console output tracing

[Steven Rostedt <rostedt@goodmis.org>]

On Thu, 2011-11-17 at 16:00 +0100, Frederic Weisbecker wrote:
> On Thu, Nov 17, 2011 at 03:57:57PM +0100, Johannes Berg wrote:
> > On Thu, 2011-11-17 at 15:55 +0100, Frederic Weisbecker wrote:
> > 
> > > > +++ b/kernel/trace/printk-trace.c	2011-11-16 21:22:20.000000000 +0100
> > > > @@ -0,0 +1,14 @@
> > > > +/*
> > > > + * printk trace points
> > > > + *
> > > > + * Copyright (C) 2011 Johannes Berg <johannes.berg@intel.com>
> > > > + */
> > > > +
> > > > +#include <linux/string.h>
> > > > +#include <linux/types.h>
> > > > +#include <linux/workqueue.h>
> > > > +#include <linux/sched.h>
> > > > +#include <linux/module.h>
> > > > +
> > > > +#define CREATE_TRACE_POINTS
> > > > +#include <trace/events/printk.h>
> > > 
> > > Why not defining the tracepoint in kernel/printk.c ?

Right, there's no reason to add another file, except to get Johannes's
copyright in the kernel ;-)

> > 
> > Is that even possible? Do you then have to do something like
> > 
> > #include <.../printk.h>
> > #define CREATE_TRACE_POINTS
> > #include <.../printk.h>
> > 
> > or how else? In any case it seemed more common to not do that.
> 
> You can just do what you did:
> 
> #define CREATE_TRACE_POINTS
> #include <trace/events/printk.h>
> 
> but in kernel/printk.c
> 
> We typically define the tracepoint from the file where an unconditional user
> resides.

The above *is* the correct way to do this.

Thanks Frederic for pointing that out. I should have spotted it, but I
blame my meds ;)

-- Steve